The performance of gypsum plasterboard assemblies is typically evaluated in accordance with standardised test methods such as BS476, AS1530 or ASTM E119. Standard time-temperature curves give good comparison between tested materials. However, they are generally less severe than a typical short duration compartment fire, they do not have a decay phase, and may be conservative for long duration fires. It is not common knowledge that test time-temperature relationships, such as ISO834, have not significantly changed since they were originally formulated in the early 1930's. Full-scale compartment testing based on typical residential scenarios conducted as part of this study, revealed that temperatures within a compartment can far exceed those of standard time-temperature curves within several minutes of ignition. Pilot-scale furnace testing to non-standard time-temperature curves has revealed that the performance of light framed gypsum plasterboard assemblies is highly dependent on the severity of the fire exposure. A system that has achieved a fire resistance rating of 60 minutes failed within 30 minutes to a fire exposure that would represent a moderate compartment fire. Current fire engineering designs often use sophisticated evacuation models to calculate minimum escape times required for safe evacuation of occupants. These evacuation times typically fall in the range from 15 - 60 minutes. The suitability of protecting escape routes using barriers rated against a standard fire test is questioned. SAPIR, a powerful finite element program, has been employed to predict the thermal behaviour of various gypsum plasterboard assemblies exposed to a range of nonstandard fires. Results from the computer modelling are compared with several full and pilot scale furnace tests. It was found that the model calibrated to results from standard ISO834 furnace testing provided reasonable predictions of temperatures within assemblies exposed to a moderate fire. Temperature predictions of assemblies exposed to severe fires were poor.